GuideMode now captures sessions from GitHub Copilot CLI, alongside Claude Code and OpenAI Codex.

If your team works across more than one coding agent, that matters. A question like “which agent makes us most productive?” can only be answered when every agent’s sessions are measured the same way. Until now, Copilot CLI was a gap in that picture.

Setting it up

Update the CLI and run setup again:

npm install -g guidemode
guidemode setup

guidemode setup looks for each agent you have installed and offers to hook it. For Copilot CLI it writes a single hooks file of its own, ~/.copilot/hooks/guidemode.json, and leaves any hooks you already have untouched. guidemode uninstall deletes that file again.

From then on, Copilot CLI sessions sync as you work, exactly as Claude Code and Codex sessions do. The upload runs in the background, so Copilot never waits on it, and your per-repository sync settings and local redaction apply unchanged.

What you get

Copilot CLI writes its own event log for every session. GuideMode converts it to the same canonical format as every other agent, so a Copilot session sits beside a Claude Code or Codex session in the same dashboards and is measured the same way:

The full transcript: your prompts, the agent’s replies, and every tool call with its result.
Subagents, recorded as part of the session that started them rather than as sessions of their own.
Token usage per model. Copilot often routes one session through several models, and each one’s usage is attributed to it.
Context window readings at the end of each turn, so you can see how full the context ran.
AI analysis: the session summary, quality assessment and phase breakdown that Claude Code and Codex sessions get.

Good to know

Copilot CLI records token usage when a session ends, not after each request. A session that is still running will show its transcript and context readings, and its token totals will appear once it closes. Resuming a session later with --resume adds only the new usage, never the same tokens twice.

Most of Copilot’s reasoning is encrypted in its own log, so GuideMode shows the reasoning summaries Copilot records, not the full chain of thought.

This covers Copilot CLI in your terminal. Copilot’s cloud coding agent and Copilot Chat in the editor are not captured.

Full details are in the GitHub Copilot CLI provider guide.